Optimizing for Value - Fight Failure Demand With Bottlenecks
summary
Argues that many "slow service" symptoms come from upstream failure demand (lots of identical redundant requests) and not raw capacity problems, uses the Theory of Constraints' five focusing steps plus concrete tactics: add observability and a Slack channel to capture failing orders, instrument and replay requests, introduce consistent error codes, fix flaky tests (took four months to reach ~90% reliable runs), and bring an ops person on call; outcome: redundant traffic falls, bottleneck shifts, and reliability and customer experience improve.
AI-written summary · about
chapters (18)
- 0:00Optimizing for value
- 1:08Failure demand and bottlenecks
- 3:00Problem definition and takeaways
- 6:33Control circles influence
- 8:17Failure demand and change
- 11:29Failure demand examples
- 13:15Discussion on failure demand
- 17:15Bottlenecks and supply/demand
- 19:03Failure in software quality
- 21:50Small steps to reduce failure
- 25:24Bottleneck process intro
- 27:09System goals and demand
- 28:39Finding bottlenecks
- 30:24Subordinating to bottleneck
- 32:12Elevating the bottleneck
- 33:16Flaky tests as bottleneck
- 36:31Vintage service case study
- 39:24Conclusion and expanding influence
related talks
29mComplexity: The Crucial Ingredient in Your Kitchen
34mDistributed Systems - What Can Go Wrong and How to Get It Right
01h 02mTools for simplified large scale systems
30mUsing the Theory of Constraints to Prioritize Applications for Migration to Cloud Foundry
28mMaking the Impossible Impossible: Improving Reliability by...
27mLessons Learned From Architecting the Highest-scale Operational Systems in the World
45mThe Map Is Not the Territory: How SLOs Lead Us Astray, and What We Can Do about It
01h 01mSoftware Architecture, Team Topologies & Complexity Science